Spatial analysis associates dECM deployment with myocardial organization changes

A retrospective analysis of two rat studies found that heart tissue exposed to decellularized extracellular matrix, or dECM, showed a directionally consistent shift toward greater spatial organization after differences among delivery methods and sampling schedules were accounted for. The researchers calculated a common tissue-level measure from 21 spatial-transcriptomic sections and estimated an average treatment-associated increase of 0.0832 on the measure’s Moran’s I scale. The estimate was positive, but its 95% confidence interval extended from −0.0240 to 0.1903, and an exact two-sided permutation test returned a P value of 0.0933. The result therefore does not establish a statistically conclusive effect, but it identifies a measurable pattern that remained positive across individual tissue fields, omitted observations, alternative spatial scales and a separate representation of tissue composition.

dECM is the protein- and glycosaminoglycan-rich scaffold left behind when cells are removed from donor tissue. Because it can preserve biochemical and structural components of the original matrix, it is being investigated as a scaffold for cardiac repair after myocardial infarction. The material in the two source studies came from porcine left-ventricular myocardium, but it was prepared in different forms. One study tested a 6 mg/mL injectable hydrogel given by a single 27-gauge intramyocardial injection; the other tested a 10 mg/mL infusible formulation administered through simulated intracoronary delivery. Processing, concentration, nominal amount, route and collection time all differed, making direct pooling of individual spatial-transcriptomic spots especially problematic. Treating each tissue section as one observation offered a way to compare these heterogeneous experiments without allowing thousands of spots within a section to be mistaken for independent animals.

The analysis combined nine publicly separable capture units from the injectable-hydrogel study with 12 sections from the infusible study, producing 79,365 reconstructed spatial spots in total. The injectable dataset included subacute tissue collected seven days after treatment at one week after infarction and chronic tissue collected seven days after treatment at eight weeks after infarction. The infusible dataset included tissue collected one, three and seven days after administration. The resulting observations fell into five study-and-time strata, each retaining its own saline-control and dECM-treated sections. This structure allowed the model to estimate a pooled dECM-minus-control difference while preserving the source studies’ different contexts. The investigators did not create new animals, perform new interventions or fabricate biomaterial; their contribution was a secondary analysis of the two public datasets.

To build the shared outcome, the researchers reconstructed each section from raw-count matrices and spatial coordinates, then normalized expression to 10,000 counts per spot. Treatment-blind single-nucleus references from matched studies and time points were used to estimate four broad fields in every spatial section: cardiomyocyte, fibroblast, myeloid and endothelial. These reference-derived fields represent estimated biological programs or cell-state signals rather than calibrated cell fractions. For each field and section, the investigators constructed a symmetrized six-nearest-neighbor graph and calculated Moran’s I, a measure of whether neighboring spots have similar values. Positive values indicate spatial clustering, values near zero indicate weak patterning relative to the chosen graph, and negative values indicate local dissimilarity. The four Moran’s I values were then averaged without weighting to create one endpoint per section, representing average within-field spatial autocorrelation rather than cell colocalization or direct cell-to-cell interaction.

In the primary stratified model, the positive coefficient was supported by directionally positive estimates for all four fields. The cardiomyocyte estimate was 0.0803, with an exact two-sided P value of 0.0766; the fibroblast estimate was 0.0770, with P = 0.0993; the myeloid estimate was 0.1410, with P = 0.1006; and the endothelial estimate was 0.0344, with P = 0.3960. Myeloid organization therefore had the largest point estimate, while the endothelial field showed the weakest association. None of these field-specific results met a conventional threshold for statistical significance, and the primary pooled result remained uncertain. The scientific importance of the analysis lies in its combination of a common calculation with strict separation of section-level replication from spot-level measurements: the thousands of spots define the spatial statistic, but the 21 public sections are the observations used for treatment inference.

The pooled direction proved stable under numerous checks. Omitting each section in turn produced coefficients from 0.0607 to 0.1093, and omitting each biological field in turn produced values from 0.0639 to 0.0994. The estimate also remained positive when the spatial graph was broadened to 18 or 36 neighbors, when a physical-distance graph based on each section’s local spot spacing was used, and when broad expression-and-state fields were constructed independently of the reference decomposition. In exploratory comparisons limited to later-sampled contexts, the dECM-minus-control contrasts were 0.1317 for chronic injectable infarction, 0.1093 for infusible delivery at day 3 and 0.1770 at day 7. These contrasts were descriptive rather than confirmatory because the subset had been motivated by patterns seen in the same data, and the analysis could not separate formulation, route, dose, injury stage or collection schedule.

Gene-level analysis offered a biological hypothesis for the spatial pattern. For each eligible rat gene, the researchers calculated section-level spatial autocorrelation, applied the same treatment-plus-stratum model and ranked genes by treatment statistics. They then mapped high-confidence one-to-one rat genes to mouse orthologs before testing preselected pathway libraries, reducing the risk that unmatched species symbols would distort enrichment. A related family of extracellular-matrix programs emerged: Gene Ontology extracellular matrix organization had a normalized enrichment score of 1.854 and a within-library false-discovery-rate-adjusted value of 0.0270, while collagen biosynthesis and modifying enzymes had a normalized score of 1.773 and an adjusted value of 0.0380. A Reactome extracellular matrix organization pathway was positive but borderline, with an adjusted value of 0.0534. Angiogenesis supplied a secondary vascular context, although its result did not remain compelling after a post hoc correction across all three libraries.

Fixed-rule pathway maps showed that the nominated extracellular-matrix program occupied structured, contiguous regions within selected control and dECM sections from both delivery platforms. These maps display where the expression program is relatively high or low within each section; they do not reveal where the administered material settled, nor do they demonstrate that matrix remodeling occurred in direct response to local dECM contact. That distinction is especially important because the same datasets were used both to select the program and to create the visual maps. The analysis therefore nominates an ECM/collagen program for prospective testing rather than demonstrating a material-driven mechanism. The pooled association could also reflect the heterogeneous source designs, broad field estimates, shared spatial or transcriptional structure, or unmeasured differences in retained tissue dose, which was unavailable in the source data.

The study’s central contribution is a common section-level axis for comparing dECM studies that use different preparations and delivery settings. It converts each spatial-transcriptomic section into a reproducible tissue-organization outcome while preserving the section as the inferential unit, and it shows that a positive direction recurs despite imprecision and sensitivity to analytic choices. Randomized follow-up experiments would need to measure material composition and mechanics before administration, then link retained dose, histology, vascular structure, spatial transcriptomics and cardiac function within the same animals. Independent tissue assays would also be needed to test whether the ECM/collagen pattern is reproducible and whether it is specific to dECM rather than a feature of infarction, remodeling or the analytical pipeline. Until then, the supported conclusion is narrow but useful: across two public rat datasets, dECM exposure was associated with a modest, directionally stable increase in a common measure of myocardial spatial organization, accompanied by pathway-level evidence pointing toward extracellular-matrix and collagen programs.

Subject of Research: Cross-study analysis of myocardial spatial organization after decellularized extracellular matrix deployment in rats

Article Title: Section-level spatial autocorrelation enables cross-study evaluation of myocardial organization after dECM deployment

Article References: Xue, Z., & Sheng, Y. (2026). Section-level spatial autocorrelation enables cross-study evaluation of myocardial organization after dECM deployment. Frontiers in Materials, 13, Article 1965824. https://doi.org/10.3389/fmats.2026.1965824

Image Credits: AI Generated

DOI: 10.3389/fmats.2026.1965824

Keywords: dECM, myocardial infarction, spatial transcriptomics, Moran’s I, cardiac repair, extracellular matrix, biomaterials, rat models, tissue organization, collagen, Section-level, spatial